[R-wiki] difficulty in editing mixed models/p-value thread

Philippe Grosjean phgrosjean at sciviews.org
Sat Dec 2 09:38:02 CET 2006 

Hello Dieter,

I read your complains too in 
http://wiki.r-project.org/rwiki/doku.php?id=guides:lmer-tests. First, 
let me tell you that I appreciate your contribution to the R Wiki and I 
seriously consider these complains.

Now, a couple of explanations. First, the Wiki IS NOT the right place to 
discuss *technical issues* of the Wiki inside pages/sections with other 
subject. Look at your page. It is about lmer... and it start with two 
paragraphs of complains about the Wiki. Doing so, you pollute your page 
yourself. The best place to discuss these technical issues is in 
R-SIG-WIKI. Ben Bolker adequately moved the discussion there. So, 
please, let's continue this discussion there; I will edit your page to 
remove these distracting elements, but I copy and paste it in the 
current message:

==============================================
 > I don’t know why this Wiki adds unwanted nonsense-links to pf(), but I
 > don’t have the patience to find out why because of the inacceptable
 > waits in editing. And it’s not my fault that MS(linux) is considered a
 > product of Microsoft by the makers of this Wiki. I also don’t use the
 > R-highlighter, because the choice of colors makes reading more
 > difficult than black-only.
 >
 > Creating the text here was not much fun, waiting times are simply
 > unacceptable and the Wiki behaves unpredictably when syntax errors are
 > detected. I still do not know why Spencer Graves’ text at the end of
 > this page is Ok in a separate link, but crashes the parser when
 > inserted in-line. This has cost me over an hour to track it down
 > because of the long waits.
 >
 > — Dieter Menne 2006/10/09
==============================================

Now, I take your various gripes one after the other, and try to give an 
answer to them:

 > I don’t know why this Wiki adds unwanted nonsense-links to pf(), but I
 > don’t have the patience to find out why because of the inacceptable
 > waits in editing.

We try hard to make Wiki pages context-sensible. The idea is: to be able 
to read a Wiki page and get instant links and hints about R functions 
cited, or used in R code sections. This is not easy to do because that 
code is not actually parsed by R, and guessing what the context is, 
sometimes fails. It is certainly possible to do better, which we will do 
in subsequent versions of the Wiki engine. However, pf() in the text is 
automatically detected as a citation to the... pf() function in package 
stats. So, the Wiki does the two following additions:

1) It creates a hint with the syntax of the function (its arguments). In 
this case, it seems correct, isn't it?

2) It is supposed to create a link to the Wiki version of the man page 
for that function. As you noticed, there is a bug for that function, and 
it incorrectly links to plot.stepfun() (sic!).

Now, when you notice that, the good behavior is to send a mail to the 
Wiki maintainer to signal the bug. So, please, do so in the future. Keep 
in mind that people are working hard to make this R Wiki, and consider 
their work: no software is totally bug-free (well, what about LaTeX?)

Now, about the slowliness you complain for. First, the Wiki edition is 
totally managed by the server, not by the client (although there is a 
little bit of JavaScript run on the client). The reason is to make it 
compatible with as many different web browsers as possible. This is hard 
to do, those days, with a client-side implementation of the editor. The 
bad side is the heavy traffic between the server and the client this 
requires for long pages with slow Internet connections. So, three 
recommendations:

1) Try to understand why the connection with the R Wiki server is so 
slow for you,

2) Please, learn how to present material in the Wiki. Your page is way 
too long and is a rather crude copy and paste from the long thread in 
the mailing list. Wiki pages are supposed to be digest forms of them, 
removing most of the discussion with section titles matching the 
content, not the name of authors of posts in the mailing list.

3) For long Wiki pages, the Wiki engines allows for editing one section 
at a time. Shorter text to edit = faster response of the editor, 
especially on slow Internet connections.


 > And it’s not my fault that MS(linux) is considered a
 > product of Microsoft by the makers of this Wiki.

This comes from the default list of acronyms provided with DokuWiki. You 
cannot deny that MS is often used as acronym for Microsoft. Now, if you 
happen to use MS with a different meaning, you can still protect it from 
the bad, automatic interpretation by writing  %%MS%%(linux). Please, 
read the Wiki syntax more carefully before complaining so badly about a 
problem. In particular, (re)read:

http://wiki.r-project.org/rwiki/doku.php?id=wiki:syntax

... and you will find there:

"To let the parser ignore an area completely (ie. do no formatting on 
it), enclose the area either with nowiki tags or even simpler, with 
double percent signs %%."

And again... if you don't find the solution yourself, you still can ask 
on R-SIG-GUI and someone would gladly give you the answer.

 > I also don’t use the R-highlighter, because the choice of colors makes
 > reading more difficult than black-only.

This is YOUR feeling. You should consider that it is not easy to find a 
combination of colors that would be usable on ANY SCREEN, not only 
yours. In particular, a good practice is to limit colors used to the 
small palette of colors that are considered "Web-safe" (same colors on 
all Web Browsers and plateforms). When you consider that, plus the fact 
that the text should remain readable (no pale yellow text on gray 
background, for instance), you end up with a VERY LIMITED choice.

Now, you are the first one to complain about color for syntax 
highlighting. So, two possibilities: (1) nobody dares to do so before 
you, (2) you are marginal in this consideration.

If its is point (1), please, propose another combination of web-safe 
colors. I have worked more than two days, trying different configs on 
different machines before chosing this one. So, I cannot do better myself.

Now, do you write the Wiki page for yourself, or for other readers. For 
other readers, of course. So, please, do follow the format and 
conventions adopted in the R Wiki. R code MUST be declared as <code 
r>...</code> and it activates also automatic links for R functions, not 
only syntax highlighting. Other readers could find black text less 
readable that colored text and would appreciate automatic links to man 
pages of R functions they don't know. At least, it is my case!

 > Creating the text here was not much fun, waiting times are simply
 > unacceptable and the Wiki behaves unpredictably when syntax errors are
 > detected. I still do not know why Spencer Graves’ text at the end of
 > this page is Ok in a separate link, but crashes the parser when
 > inserted in-line. This has cost me over an hour to track it down
 > because of the long waits.

Again, please, write first the structure of your page (the section 
titles) and then, edit these sections one at a time if your Internet 
connection is slow. Also, read more carefully the syntax of the Wiki. 
You complain about bad formatting of your pages, but the Wiki engine is 
just a robot that does what YOU tell him to do. If you give bad 
instructions, you are guilty, not the Wiki engine! This will take me 
MORE than one hour to both answer your complains and reformat your page 
correctly... and it is certainly not much fun for me either!

Now, I will try to edit your page to present it the way it should be in 
the Wiki. Would you consider the work of other people a little bit more 
positively, would you formulate bug reports or suggestions the right 
way, and you would be really appreciated as a R Wiki contributor.
Best,

Philippe Grosjean

Ben Bolker wrote:
>    Dieter,
> 
>   I read your complaints about waiting times and
> difficulties with the parser in constructing
> http://wiki.r-project.org/rwiki/doku.php?id=guides:lmer-tests
> 
>   I didn't seem to have any trouble -- could it somehow
> be something platform- or connection-specific?
> (I'm using Firefox 1.5.0.8, Ubuntu 6.06) Or some transient
> problem?
> 
>   I agree it would be nice to figure out why pf() links
> to plot.stepfun(stats) ??  And to turn off the MS = Microsoft
> tooltip ...
>   I'm not sure I agree about the syntax coloring being
> unhelpful ...
> 
>    Ben Bolker
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> R-sig-wiki mailing list
> R-sig-wiki at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-wiki

More information about the R-sig-wiki mailing list